Birders and ornithologists, listen up. Using four large data sets of recorded bird calls, scientists at Queen Mary University in London have you record with previously recorded bird calls in its database.
The researchers, who in the journal PeerJ, wanted to tap into the growing accuracy of feature learning—a computational input method that takes foreign raw data and compares it automatically to information already in a system, rather than using prelabeled data. Their technology enables birders, who before may have had to resort to tedious deductive research to identify birds in their area, to upload information of their own, much like popular music-identifying application .
To test the idea, the researchers entered their creation in a public contest of identifying more than 500 Brazilian birds’ calls from thousands of recordings. The tech placed second out of 10 programs, and it won the “audio only” category.
The researchers have higher hopes for this technology than just helping you avoid awkward call-and-response-style identification of backyard birds. Lead researcher Dr. Dan Stowell, who received a grant for this project, says it might give insight into the evolution of human language and social organization.
“Birdsong has a lot in common with human language, even though it evolved separately. For example, many songbirds go through similar stages of vocal learning as we do as they grow up, which makes them interesting to study,” he said. “The attraction of fully automatic analysis is that we can create a really large evidence base to address these big questions.”
Up next: Stowell is working on techniques to distill even more information from raw song recordings that could revolutionize our understanding of birds’ social worlds. In addition to identifying birds, the tech might be able to analyze longer recordings of multiple birds in a scene, telling you which bird was talking when, which birds responded to it, what their relationship is like, and what that tells you about who’s leading the conversation.